Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2025 Aug 1.
Published in final edited form as: Int J Obes (Lond). 2024 Oct 29;49(2):322–331. doi: 10.1038/s41366-024-01665-6

Shaping childhood obesity: behavioral and environmental risk factors associated with body mass index trajectories between 2 and 9 years in Samoan children

Courtney C Choy 1,2, William Johnson 3,*, Take Naseri 4,2, Vaimoana Filipo 5, Maria Siulepa Arorae 5, Faatali Tafunaina 5, Folla Unasa 5, Kima Savusa 5, Muagututia S Reupena 6, Joseph M Braun 2, Rachel L Duckham 7,8, Christina Soti-Ulberg 4, Stephen T McGarvey 2,9,*, Nicola L Hawley 1,*
PMCID: PMC11805637  NIHMSID: NIHMS2035568  PMID: 39472691

Abstract

Background/Objective:

Pacific children are at high obesity risk, yet the behavioral and environmental factors that contribute to obesity development in this setting remain poorly understood. We assessed associations between childhood risk factors for obesity with body mass index (BMI) trajectories between ages 2-9 years in Samoa.

Subjects/Methods:

In a prospective cohort of 485 children from ‘Upolu, we measured weight and height at ages 2-4 (2015), 3.5-8 (2017-18), and 5.5-11 years (2019-20). Modern dietary pattern adherence was assessed using factor analysis of primary caregiver-reported food frequency questionnaire data. Physical activity was estimated with the Netherlands Physical Activity Questionnaire. Socioeconomic resources were assessed using an 18-item household asset score. Urbanicity was based on village residence. Associations of these risk factors with predicted weight, height, and BMI (at 1-year intervals and velocity) were assessed using multilevel cubic spline regressions.

Results:

Females had greater adjusted weight velocity with high modern dietary pattern adherence compared to low (p-value for interaction with age spline term 1=0.028 and age spline term 2=0.007). Starting at age 3 years, children with higher physical activity had higher BMI, but this association was not meaningful up to age 9 (all p-value>0.05). Males with very high compared to low household assets had higher BMI from age 2 to 4 years (95% CI: 0.26-1.53 kg/m2, p=0.006) and greater BMI velocity (p-value for interaction with age spline term 2=0.001). Males in the urban region had the greatest BMI gain after age 5 compared to the rural region (p-value for interaction with age spline term 2=0.014).

Conclusions:

High, centile-crossing BMI trajectories suggest that obesity prevention and intervention are needed among Samoan children before age 9 years. Positive associations between high modern dietary pattern adherence, greater asset ownership, and urbanization offer initial insights into who, and which behavioral risk factors, should be prioritized in implementing public health solutions.

Introduction

Despite considerable investments in the prevention and treatment of obesity and related cardiometabolic diseases in children, their prevalence continues to increase globally with catastrophic social and economic costs predicted if the epidemic is not halted (1). More effective strategies for prevention are urgently needed and current intervention approaches must be optimized with better identification of the most efficacious targets and timing. In Samoa, an independent country in Polynesia, children are among the most at risk for obesity globally (2). However, the majority of obesity-focused health research among Samoans has been cross-sectional and studies have generally not started early enough in childhood to understand behaviors and environmental risk factors that may shape body size trajectories and place children at risk for overweight or obesity.

The Samoan Ola Tuputupua’e “Growing Up” study is the first child cohort among the independent Pacific Island countries with contemporary, longitudinal data to characterize changes in body size during early and mid-childhood - a critical period for obesity development. Cross-sectional data from the cohort shows an increasing prevalence of overweight and obesity from 16% among children at ages 2-4 years (3), to 25% at 3.5-7 years (4), and 36.2% at 5.5-11 years (5). Among children at ages 2-4 years, those who lived in a household with a higher number of consumer durable assets had a higher odds of overweight or obesity, highlighting the potential role of the home environment and socioeconomic resources in obesity development and body size in childhood. Children who had consistently high adherence to a modern dietary pattern between the ages of 2 and7 years had greater gains in WHO z-scores for weight and BMI (4) and tended to reside in the urban rather than rural region (6, 7), which aligns with nutrition transition-related obesity trends in Samoa (7, 8). Patterns of physical activity, and their role in obesity risk, remain poorly understood in this setting. While maternal-reported physical activity was not cross-sectionally associated with overweight and obesity prevalence among children at ages 2-4 years (3), more time spent in moderate-vigorous physical activity (as measured by accelerometry) was associated with greater lean mass in males, but not females at ages 3.5-7 years (9).

To inform prevention efforts we sought to build upon this prior work to understand the extent to which changes in body size in childhood are shaped by these modifiable behaviors and environmental risk factors for obesity. The objectives were to examine associations between childhood dietary patterns and physical activity (as modifiable behavioral risk factors) and household socioeconomic resources and urbanicity (as structural environmental risk factors) with BMI trajectories between ages 2-9 years in the Ola Tuputupua’e cohort.

Methods

Study Design, Subjects, and Data Collection

The mixed-longitudinal Ola Tuputupua’e study includes 501 children with serial measurements across three data collection waves in 2015 (at ages 2-4 years), 2017-18 (3.5-8 years), and 2019-20 (5.5-11 years). A convenience sample of 319 Samoan children and their mothers were initially recruited in 2015 from ‘Upolu island villages to ensure equal representation from census regions with varying levels of urbanicity and exposure to nutrition transition: Rest of ‘Upolu (ROU; rural, low exposure), Northwest ‘Upolu (NWU; periurban, moderate exposure) and Apia Urban Area (AUA, urban, high exposure) (3). In 2017-18, we followed up with 84.6% of children from 2015 (n=277; 3.5-8 years) and recruited additional eligible children (n=182) to expand the cohort and include one additional AUA village. In 2019-20, at the third data collection wave, the retention rate was 87.4% (5).

Ethics approval and consent to participate

Yale and Brown University Institutional Review Boards and the Health Research Committee of the Samoa Ministry of Health (IRB# 2000020519; IAA# 18-41 959) approved protocols for each data collection wave. Parents provided written informed consent at every wave and children over the age of seven gave their written assent. All methods were performed in accordance with the Declaration of Helsinki.

Study sample

The sample included children with complete data from any data collection wave (2015, 2017-18, and/or 2019-20). We chose to restrict this analysis to ages 2-9 years because of the limited number of children who were age 10 (n=11) and age 11 (n=4). Of the 501 children in the cohort, 5 were outside the eligible age range, and 11 had missing or implausible data. The final sample was 485 (Supplementary Figure 1). The majority (n=437, 90%) had ≥2 measurements (range: 1-3), and a total of 1,177 weight and height measurements were used.

Outcome Assessments

Childhood BMI trajectories were the primary outcomes of interest; weight and height are presented as secondary outcomes. Standardized procedures were used to measure weight (HD 351 weight scale; Tanita Corporation of America, Arlington Heights, IL) and standing height (Stadiometer Pfister Imports, New York, NY or Seca, CA, USA) (both in duplicate ±0.1 kg and ±0.5 cm). We chose to model trajectories using average weight (kg), height (cm), and BMI (kg/m2) because these biological traits are able to better capture the variability and changes in childhood body size compared to Z-scores (10).

Exposure Assessments

Child dietary patterns, physical activity, household socioeconomic resources, and urbanicity were the exposures of interest. We chose to focus on the assessment of each behavior and environmental factor at the time of enrollment (when median child age was 4 years, in either 2015 or 2017) to explore how these factors are prospectively associated with body size trajectories across childhood.

A ‘modern’ dietary pattern (named to distinguish this pattern from the more traditional island diet) was previously identified among the cohort using maternal-reported food frequency questionnaire (FFQ) data and principal component analysis (4, 7). This FFQ was validated in Samoan adults (11–13) and adapted for use in the cohort. The modern dietary pattern included high intakes of french fries, unprocessed red meat, potatoes/sweet potatoes, cereals, noodles, and fruit juices, and low intakes of breadfruit and taro (traditional local starches). Similar to another study (14), pattern factor scores were dichotomized as “low” or “high” according to the sample median to capture each child’s relative degree of adherence to the modern dietary pattern.

Physical activity was assessed using 3-items of the Netherlands Physical Activity Questionnaire for young children (15–17). In a validation study, among a sub-sample of the cohort who wore Actigraph GT3X+ accelerometers in 2015, we reduced the questionnaire from 7 to 3 items (with a maximum score of 15) that provided the greatest internal consistency (18). Scores were categorized into tertiles to distinguish low, medium, and high levels of maternal-reported activity.

Household socioeconomic resources were assessed using an 18-item asset score, an approach previously used among Samoan adults (6), which had a strong positive association with overweight and obesity prevalence among children at ages 2-4 years in Samoa (3). Asset scores were categorized into quartiles based on distribution to categorize participant households as having low, medium, high, and very high assets and therefore socioeconomic resources. Urbanicity (rural ROU, peri-urban NWU, and urban AUA) was based on village census region where the child resided (3).

Covariate Assessments

All covariates were measured at enrollment (in 2015 or 2017) and were selected a priori for inclusion in each multilevel model to minimize potential sources of selection bias and confounding.

Child age in years was calculated by subtracting reported date of birth from the date of any survey and physical assessment. Mothers recalled whether they ever breastfed or gave their child their pumped milk, responding “yes” or “no”. Children’s daily total energy intake was estimated by multiplying frequency of each food reported in the FFQ (times per day) by the nutrient content of fixed, standard portion size. Mothers reported their age in years, highest level of educational attainment, marital status, and total annual income of the household, using questions similar to those in census surveys in Samoa (19, 20). To enhance interpretability of the results, we maintained the same categories when reporting covariate data as previously published for the cohort (3, 4). Maternal weight and height were measured using the same protocols described for children and used to calculate BMI.

Guided by a directed acyclic graph (21) and consistent with prior research (22–26), we identified and adjusted for all potential measured confounders, while avoiding inappropriate adjustment for potential mediators to minimize bias in our assessment of the relationship between each exposure (child dietary patterns, physical activity, household socioeconomic resources, and urbanicity) and BMI trajectories. We considered child dietary patterns and physical activity to be potential mediators in relationships between structural environmental risk factors and body size trajectories and thus, did not adjust for these behavioral risk factors in models with household assets or census region as the main exposure. Similarly, relationships between urbanicity and body size trajectories may be mediated by family socioeconomic resources and economic development (27–29), so maternal characteristics and household assets were not adjusted for in models with census region as the exposure.

Statistical Analyses

We first summarized sample characteristics at enrollment. We present numbers and percentages for categorical variables, means, and standard deviations (SDs) for normally-distributed continuous variables (Kolmogorov-Smirnov goodness-of-fit test, p>0.05), and medians with lower (Q1) and upper (Q3) quartiles for non-normally-distributed continuous variables. To assess differences in sample characteristics by sex we used Chi-square tests, generalized linear models, or exact Wilcoxon two-sample tests (30).

Multilevel models were used to analyze longitudinal data, as previously described (31). Trajectories for serial BMI, weight, and height were modeled with measurement occasion at level 1 and individuals at level 2. Models were sex-stratified because obesity risk factors, particularly behavioral risk factors, are known to vary by sex in this setting (32) and their associations with body size trajectories may differ. The shape of each body size trajectory was specified as restricted cubic spline age terms, with 3 knots for BMI and weight (age centered at 2 years: 0.90, 3.72, 6.69) and 4 knots for height (0.40, 2.68, 4.84, 7.24) because this provided the best model fit with lowest Bayesian information criteria (BIC). Diagnostic plots verified that normality and homoscedasticity assumptions were met. We used an unstructured variance-covariance matrix among random effects at each level so that each term is estimated and not constrained. In this way, the intercept (e.g., mean BMI at age 2 years) and slope (e.g., BMI velocity between age 2 and 9 years) are allowed to vary for each child.

Separate models for each exposure were fitted to assess their association with mean body size trajectories (Supplementary File 1). For each model, child age was centered at different values between 2 and 9 years to improve the interpretability of the coefficients. Each exposure was entered as a main effect and an interaction with the age spline terms to allow for the mean trajectories to differ across each level of exposure (e.g. low, medium, high, or very high levels of household assets) and assess their associations. All covariates were added as time-fixed, main effects in the model.

For data visualization, linear predictions from these models were converted to z-scores (SD and centile) and plotted to illustrate trajectories of body size (weight, height, or BMI) relative to the WHO standards from age 2 up to age 5 and WHO references from age 5 to 9. We further used fully adjusted models to estimate mean differences in body size between exposure groups at one-year age intervals and assess associations.

SAS 9.4. (SAS Institute, NC, USA) and STATA 16.0 (StataCorp, TX, USA) were used for analyses and a two-sided α of 0.05 was specified. When relevant, point estimates and 95% confidence intervals (CI) were estimated. Findings were interpreted in line with hypothesis testing literature (32).

Results

Enrollment characteristics at age 2 to 5 years

Among the 485 children, there were no sex differences in modern dietary pattern adherence, household assets, and urbanicity at the time of enrollment in 2015 and 2017 (Table 1). Females had lower reported physical activity than males (21.77% versus 32.07% were in the highest tertile, p=0.028). Despite similar weight and height, females had lower median BMI (16.51 kg/m2, Q1-Q3: 15.79 – 17.63) than males (median: 16.96 kg/m2, Q1-Q3: 15.96 – 17.82, p=0.038). On average, children remained above the median weight and BMI for the WHO child standard and reference groups (Supplementary Figure 2).

Table 1.

Sample characteristics by sex at initial recruitment in the Ola Tuputupua’e study, 2015-2017†

Total (N=485) Female (n=248) Male (n=237) P‡
Individual
Age, years 4.07 (2.92 – 4.96) 4.10 (2.93 – 5.01) 4.06 (2.92 – 4.94) 0.836
High adherence to modern diet 245 (50.52) 120 (48.39) 125 (52.74) 0.338
Physical activity score tertile§: Low 211 (43.51) 119 (47.98) 92 (38.82) 0.028
  Medium 144 (29.69) 75 (30.24) 69 (29.11)
  High 130 (26.80) 54 (21.77) 76 (32.07)
Total energy intake, 1000 cal/day 2.68 (1.63 – 5.17) 2.76 (1.58 – 5.57) 2.65 (1.70 – 4.66) 0.600
Ever breastfed 351 (72.37) 185 (74.60) 166 (70.04) 0.262
Screentime, min/day 61 (60 –120) 67.50 (60 –120) 60 (60 –120) 0.876
Nighttime sleep, hr/day 9 (9 – 10) 9 (9 – 10) 9 (9 – 10) 0.749
Body size
  Weight, kg 16.70 (14.10 – 19.50) 16.00 (14.03 – 19.48) 17.10 (14.20 – 19.60) 0.318
  Height, cm 100.21 ± 11.78 100.13 ± 11.97 100.30 ± 11.60 0.874
  Body mass index, kg/m2 16.71 (15.86 –17.72) 16.51 (15.79 – 17.63) 16.96 (15.96 – 17.82) 0.038
WHO Z-scores
  Weight 0.22 (−0.48 – 0.86) 0.26 (−0.38 – 0.77) 0.21 (−0.52 – 0.95) 0.929
  Height −0.62 (−1.35 – 0.14) −0.57 (−1.21 – 0.15) −0.68 (−1.59 – 0.13) 0.122
  BMI 0.94 (0.37 – 1.57) 0.78 (0.28 – 1.45) 1.04 (0.50 – 1.64) 0.010
Maternal
Age group: 18-24 years 81 (16.70) 39 (15.73) 42 (17.72) 0.839
  25-39 272 (56.08) 141 (56.85) 131 (55.27)
  ≥40 132 (27.22) 68 (27.42) 64 (27.00)
Education, years 12 (12-13) 12 (12-13) 12 (12-13) 0.921
Married or cohabitating** 383 (78.97) 200 (80.65) 183 (77.22) 0.354
Body size
  Weight, kg 86.55 (74.85 – 99.60) 87.55 (76.53 – 104.28) 86.30 (73.55 – 97.35) 0.060
  Height, cm 161.47 ± 5.27 161.47 ± 5.09 161.46 ± 5.46 0.980
  Body mass index, kg/m2 33.35 (29.11 – 37.86) 33.78 (30.09 – 38.90) 33.11 (28.35 – 37.17) 0.029
Household
Asset score quartile†† : Low 100 (20.62) 56 (22.58) 44 (18.57) 0.378
  Medium 106 (21.86) 53 (21.37) 53 (22.36)
  High 152 (31.34) 70 (28.23) 82 (34.60)
  Very High 127 (26.19) 69 (27.82) 58 (24.47)
Urbanicity-Region: 0.482
  Rural-Rest of Upolu 149 (30.72) 72 (29.03) 77 (32.49)
  Periurban-Northwest Upolu 170 (35.05) 85 (34.27) 85 (35.86)
  Urban-Apia Urban Area 166 (34.23) 91 (36.69) 75 (31.65)
Annual income ≥$10,000 WST 93 (19.42) 48 (19.67) 45 (19.15) 0.885
Open in a new tab

WHO, World Health Organization; WST, Western Samoan tala; SD, standard deviation; P, Probability estimat

†

Characteristics were measured when child was first recruited into the cohort in either 2015 or 2017. Medians and interquartiles in parentheses (25th – 75th percentiles) are presented for continuous variables that are not normally distributed (Kolmogorov-Smirnov goodness of fit test, p ≤0.05). Means ± SDs are presented for continuous variables that are normally distributed (Kolmogorov-Smirnov goodness of fit test, p>0.05). Counts and percentages in parentheses are presented for categorical variables. Percentages may not sum to 100% due to rounding

‡

Exact p-values or Monte-Carlo estimation of exact p-values were based on the exact Wilcoxon two-sample tests for continuous variables that are not normally distributed to assess differences by sex. P values are also based on two-sample t tests for continuous variables that are normally distributed and chi-square test for categorical variables.

§

The sum of 3-items from the Netherlands Physical Activity Questionnaire for young children. The score is out of a maximum of 15, with higher scores indicating greater physical activity, and were classified into tertiles: low (mean: 9.53, SD: 2.25), medium (mean: 13.15, SD: 0.36), high (mean: 15.00, SD: 0).

**

Compared to mothers who reported being separated, divorced, or never married.

††

The sum of consumer durables owned (fridge, freezer, stereo, portable stereo, microwave oven, rice cooker, blender, sewing machine, television, VCR/DVD, couch, washing machine, landline telephone, computer/laptop, tablet, electric fan, air conditioner, and motor vehicle). The score is out of a maximum of 18, with higher scores indicating greater socioeconomic resources, and was classified into quartiles: low (mean:1.17, SD: 0.80), medium (mean: 3.52, SD: 0.50), high (mean: 5.94, SD: 0.80), and very high (mean: 10.90, SD: 2.85).

Associations of childhood modern dietary pattern adherence with body size trajectories

High adherence to a modern dietary pattern at enrollment was associated with higher childhood BMI and velocity between ages 2-9 years compared to low adherence (Figures 1–2), after adjusting for covariates. Children with high adherence experienced an increasing BMI between ages 5-9 years that crossed above the 90th centile (33), indicating a relatively high BMI for age in this group compared to the WHO child reference group and on average, were classified with overweight or obesity (BMI z-score > +1 SD). While females with high adherence at enrollment had greater weight velocity compared to low (p-value for interaction with age spline term 1 = 0.028 and age spline term 2 = 0.007, Supplementary Table 1 and Supplementary Figure 3), this association was weaker in males (p-value for interaction with age spline term 1 = 0.574 and age spline term 2 = 0.009, Supplementary Table 1). There were no differences in height trajectories by modern dietary pattern adherence groups (Supplementary Table 1).

Figure 1. Female BMI Z-score trajectories and centiles between 2 and 9 years of age according to modifiable behaviors and environmental factors.

Figure 1.

Open in a new tab

Trajectories were estimated from 248 Samoan fmales in multilevel models with (A) modern dietary pattern adherence, (B) physical activity tertile, (C) household asset quartile, and (D) census region as exposures of interest. Estimated BMI z-scores and centiles were based on the WHO child growth standards for ages 2-4 years and WHO references for ages 5-9 years. Models were adjusted for total energy intake, ever breastfed status, maternal age group, maternal education, maternal height, maternal BMI, household total annual income, when appropriate, and are presented in Supplementary Tables 1–4.

Figure 2. Male BMI Z-score trajectories and centiles between 2 and 9 years of age according to modifiable behaviors and environmental factors.

Figure 2.

Open in a new tab

Trajectories were estimated from 237 Samoan males in multilevel models with (A) modern dietary pattern adherence, (B) physical activity tertile, (C) household asset quartile, and (D) census region as exposures of interest. Estimated BMI z-scores and centiles were based on the WHO child growth standards for ages 2-4 years and WHO references for ages 5-9 years. Models were adjusted for total energy intake, ever breastfed status, maternal age group, maternal education, maternal height, maternal BMI, household total annual income, when appropriate, and are presented in Supplementary Tables 1–4.

Associations of childhood physical activity with body size trajectories

Positive associations were observed between reported physical activity and childhood body size, with the highest BMI in the high physical activity tertile (Figures 1–2). Comparing the high versus low tertiles, a BMI difference of 0.48 kg/m2 was observed in females at age 4 years (95% CI: 0.03-0.93), but not at age 9 years (0.56 kg/m2, 95% CI: −0.95 - 2.08) (Table 2). Males with high physical activity had a greater BMI than those with low activity at ages 3 to 8 years, where a large difference in BMI of 1.50 kg/m2 was observed (95% CI: 0.09 – 2.91). There was no association between physical activity tertile and BMI velocity (p-value for interaction with age spline terms >0.05, Supplementary Table 2).

Table 2.

Mean differences in childhood weight, height, and BMI by modern dietary pattern adherence, physical activity tertiles, household asset quartiles, and urbanicity

Age, years
Female (n=248)
Male (n=237)
Weight, kg
 Height, cm
 BMI, kg/m2
 Weight, kg
 Height, cm
 BMI, kg/m2
Est LL UL P Est LL UL P Est LL UL P Est LL UL P Est LL UL P Est LL UL P
High modern dietary pattern adherence vs. low† 2 1.07 0.29 1.86 0.008 −0.01 −1.31 1.30 0.993 0.40 −0.10 0.90 0.119 0.88 −0.19 1.94 0.106 1.15 −0.15 2.45 0.083 0.12 −0.49 0.74 0.700
3 0.67 0.16 1.18 0.011 0.09 −0.63 0.81 0.807 0.24 −0.09 0.57 0.161 0.74 0.06 1.42 0.034 0.72 −0.02 1.46 0.057 0.16 −0.24 0.56 0.431
4 0.27 −0.14 0.67 0.201 0.19 −0.49 0.86 0.591 0.07 −0.17 0.32 0.555 0.60 0.10 1.10 0.018 0.29 −0.34 0.92 0.370 0.20 −0.09 0.49 0.168
5 −0.14 −0.71 0.44 0.639 0.28 −0.95 1.51 0.654 −0.09 −0.42 0.24 0.598 0.47 −0.23 1.16 0.186 −0.14 −1.25 0.97 0.804 0.24 −0.13 0.62 0.207
6 −0.54 −1.40 0.32 0.220 0.38 −1.55 2.30 0.700 −0.25 −0.75 0.25 0.322 0.33 −0.75 1.41 0.545 −0.57 −2.32 1.18 0.522 0.28 −0.30 0.87 0.341
7 −0.94 −2.13 0.25 0.121 0.47 −2.17 3.12 0.725 −0.41 −1.11 0.28 0.242 0.20 −1.32 1.71 0.799 −1.00 −3.42 1.42 0.418 0.32 −0.50 1.15 0.440
8 −1.35 −2.88 0.19 0.086 0.57 −2.81 3.95 0.741 −0.58 −1.48 0.32 0.323 0.06 −1.90 2.03 0.951 −1.43 −4.54 1.68 0.367 0.36 −0.71 1.44 0.506
9 −1.75 −3.63 0.13 0.069 0.67 −3.45 4.79 0.751 −0.74 −1.85 0.37 0.192 −0.08 −2.50 2.35 0.952 −1.86 −5.66 1.94 0.337 0.40 −0.92 1.73 0.550
Medium physical activity vs. low‡ 2 0.10 −0.88 1.08 0.846 −0.86 −2.85 1.14 0.401 0.19 −0.47 0.86 0.570 0.88 −0.59 2.36 0.239 0.41 −1.74 2.57 0.706 0.40 −0.42 1.22 0.334
3 0.30 −0.40 1.00 0.403 −0.53 −2.09 1.03 0.505 0.34 −0.14 0.83 0.168 0.75 −0.21 1.71 0.127 0.37 −1.37 2.11 0.679 0.40 −0.17 0.97 0.166
4 0.50 −0.22 1.22 0.170 −0.20 −1.68 1.28 0.786 0.50 0.07 0.92 0.022 0.61 −0.21 1.43 0.144 0.32 −1.34 1.98 0.704 0.40 −0.07 0.86 0.092
5 0.70 −0.31 1.72 0.174 0.12 −1.69 1.93 0.896 0.65 0.13 1.16 0.014 0.47 −0.72 1.66 0.437 0.27 −1.67 2.22 0.783 0.40 −0.20 0.99 0.190
6 0.91 −0.52 2.33 0.213 0.45 −1.94 2.83 0.713 0.80 0.09 1.51 0.026 0.33 −1.44 2.11 0.713 0.23 −2.26 2.71 0.858 0.39 −0.46 1.25 0.366
7 1.11 −0.77 2.98 0.247 0.77 −2.29 3.84 0.622 0.95 0.02 1.89 0.046 0.20 −2.23 2.62 0.875 0.18 −2.97 3.33 0.911 0.39 −0.77 1.55 0.509
8 1.31 −1.04 3.66 0.274 1.10 −2.70 4.90 0.571 1.10 −0.08 2.28 0.067 0.06 −3.04 3.16 0.971 0.13 −3.74 4.00 0.946 0.39 −1.10 1.88 0.608
9 1.51 −1.31 4.34 0.294 1.42 −3.14 5.98 0.541 1.25 −0.18 2.69 0.087 −0.08 −3.87 3.70 0.967 0.09 −4.54 4.71 0.971 0.39 −1.44 2.21 0.678
High physical activity vs. low‡ 2 0.07 −0.89 1.03 0.888 −0.42 −2.40 1.57 0.681 0.44 −0.20 1.09 0.180 0.35 −0.99 1.68 0.611 0.13 −1.82 2.09 0.895 0.25 −0.49 0.98 0.512
3 0.29 −0.41 0.99 0.412 −0.09 −1.70 1.53 0.916 0.46 −0.03 0.95 0.064 0.66 −0.19 1.51 0.127 0.07 −1.54 1.68 0.932 0.45 −0.05 0.96 0.078
4 0.51 −0.26 1.29 0.193 0.24 −1.34 1.83 0.765 0.48 0.03 0.93 0.038 0.98 0.23 1.73 0.011 0.01 −1.55 1.57 0.990 0.66 0.24 1.09 0.002
5 0.74 −0.39 1.86 0.199 0.57 −1.33 2.48 0.557 0.50 −0.07 1.06 0.087 1.30 0.16 2.44 0.025 −0.05 −1.88 1.78 0.956 0.87 0.31 1.44 0.002
6 0.96 −0.61 2.53 0.232 0.90 −1.54 3.34 0.470 0.51 −0.26 1.28 0.191 1.62 −0.09 3.32 0.063 −0.11 −2.41 2.19 0.924 1.08 0.27 1.90 0.009
7 1.18 −0.88 3.24 0.261 1.23 −1.85 4.31 0.434 0.53 −0.48 1.53 0.302 1.93 −0.39 4.26 0.104 −0.17 −3.06 2.71 0.906 1.29 0.19 2.39 0.022
8 1.40 −1.16 3.96 0.283 1.56 −2.22 5.34 0.419 0.55 −0.71 1.80 0.394 2.25 −0.72 5.22 0.137 −0.23 −3.75 3.29 0.896 1.50 0.09 2.91 0.037
9 1.63 −1.44 4.70 0.299 1.89 −2.61 6.39 0.411 0.56 −0.95 2.08 0.467 2.57 −1.05 6.19 0.164 −0.30 −4.48 3.89 0.890 1.71 −0.01 3.43 0.051
Medium asset quartile vs. low§ 2 0.52 −0.65 1.70 0.382 −0.05 −2.37 2.27 0.969 0.41 −0.39 1.21 0.316 0.45 −1.24 2.14 0.603 −0.73 −3.34 1.88 0.584 0.74 −0.22 1.71 0.131
3 0.21 −0.65 1.06 0.637 −0.49 −2.36 1.38 0.610 0.30 −0.30 0.90 0.328 0.40 −0.73 1.52 0.489 −0.13 −2.29 2.04 0.908 0.52 −0.16 1.21 0.132
4 −0.11 −1.02 0.80 0.808 −0.93 −2.74 0.88 0.316 0.19 −0.36 0.73 0.495 0.34 −0.70 1.39 0.520 0.47 −1.63 2.58 0.659 0.30 −0.28 0.89 0.311
5 −0.43 −1.72 0.86 0.513 −1.37 −3.55 0.81 0.218 0.08 −0.59 0.75 0.817 0.29 −1.25 1.83 0.712 1.08 −1.38 3.53 0.390 0.09 −0.67 0.84 0.824
6 −0.75 −2.55 1.05 0.416 −1.81 −4.62 1.00 0.207 −0.03 −0.94 0.88 0.946 0.24 −2.02 2.49 0.837 1.68 −1.39 4.75 0.285 −0.13 −1.20 0.94 0.807
7 −1.07 −3.43 1.30 0.376 −2.25 −5.82 1.32 0.216 −0.14 −1.33 1.05 0.816 0.18 −2.86 3.22 0.906 2.28 −1.56 6.11 0.244 −0.35 −1.79 1.08 0.631
8 −1.38 −4.33 1.56 0.357 −2.69 −7.07 1.69 0.229 −0.25 −1.74 1.24 0.740 0.13 −3.72 3.98 0.947 2.88 −1.79 7.55 0.227 −0.57 −2.39 1.25 0.539
9 −1.70 −5.24 1.83 0.345 −3.13 −8.37 2.10 0.241 −0.36 −2.17 1.44 0.694 0.08 −4.60 4.75 0.975 3.48 −2.06 9.03 0.219 −0.79 −3.01 1.43 0.486
High household asset quartile vs. low§ 2 1.07 −0.01 2.15 0.052 0.73 −1.42 2.89 0.504 0.61 −0.13 1.35 0.107 0.53 −1.06 2.12 0.516 0.53 −1.93 3.00 0.672 0.64 −0.27 1.55 0.170
3 0.90 0.10 1.70 0.027 0.75 −1.01 2.52 0.404 0.51 −0.05 1.07 0.074 0.57 −0.49 1.63 0.295 0.37 −1.66 2.41 0.721 0.53 −0.12 1.17 0.108
4 0.73 −0.13 1.59 0.095 0.77 −0.96 2.50 0.384 0.41 −0.10 0.93 0.116 0.61 −0.37 1.58 0.224 0.21 −1.76 2.18 0.835 0.42 −0.13 0.97 0.136
5 0.56 −0.65 1.77 0.365 0.78 −1.28 2.85 0.458 0.32 −0.31 0.95 0.325 0.65 −0.78 2.07 0.375 0.05 −2.25 2.35 0.967 0.31 −0.39 1.02 0.385
6 0.39 −1.29 2.07 0.650 0.80 −1.84 3.44 0.553 0.22 −0.63 1.06 0.611 0.68 −1.40 2.77 0.520 −0.11 −3.00 2.78 0.939 0.21 −0.79 1.20 0.686
7 0.22 −1.98 2.41 0.845 0.82 −2.51 4.15 0.631 0.12 −0.98 1.22 0.828 0.72 −2.09 3.53 0.614 −0.27 −3.89 3.34 0.882 0.10 −1.24 1.44 0.886
8 0.05 −2.68 2.78 0.972 0.83 −3.24 4.91 0.688 0.02 −1.35 1.40 0.973 0.76 −2.80 4.32 0.675 −0.44 −4.85 3.98 0.846 −0.01 −1.71 1.69 0.991
9 −0.12 −3.39 3.15 0.942 0.85 −4.00 5.70 0.731 −0.07 −1.73 1.58 0.931 0.80 −3.53 5.13 0.717 −0.60 −5.85 4.66 0.823 −0.12 −2.19 1.95 0.912
Very high asset quartile vs. low§ 2 1.34 0.13 2.54 0.029 2.08 −0.31 4.46 0.088 0.53 −0.30 1.35 0.210 1.98 0.25 3.71 0.025 2.12 −0.61 4.84 0.128 1.32 0.33 2.31 0.009
3 0.98 0.07 1.89 0.036 0.98 −1.00 2.96 0.331 0.53 −0.12 1.17 0.108 1.88 0.69 3.08 0.002 2.81 0.48 5.13 0.018 1.11 0.38 1.83 0.003
4 0.62 −0.32 1.56 0.198 −0.12 −2.03 1.80 0.905 0.52 −0.06 1.10 0.078 1.79 0.67 2.91 0.002 3.49 1.22 5.76 0.003 0.90 0.26 1.53 0.006
5 0.26 −1.01 1.52 0.689 −1.21 −3.44 1.01 0.285 0.52 −0.16 1.20 0.134 1.70 0.12 3.27 0.035 4.18 1.59 6.77 0.002 0.68 −0.11 1.48 0.091
6 −0.10 −1.83 1.63 0.908 −2.31 −5.09 0.47 0.104 0.52 −0.37 1.40 0.252 1.60 −0.66 3.86 0.165 4.87 1.70 8.04 0.003 0.47 −0.62 1.56 0.395
7 −0.46 −2.70 1.78 0.687 −3.41 −6.88 0.07 0.055 0.52 −0.63 1.66 0.377 1.51 −1.51 4.53 0.328 5.56 1.66 9.45 0.005 0.26 −1.18 1.70 0.723
8 −0.82 −3.60 1.96 0.564 −4.50 −8.74 −0.27 0.037 0.51 −0.91 1.94 0.480 1.41 −2.40 5.23 0.468 6.24 1.54 10.95 0.009 0.05 −1.77 1.87 0.958
9 −1.18 −4.52 2.16 0.488 −5.60 −10.63 −0.57 0.029 0.51 −1.20 2.22 0.560 1.32 −3.31 5.95 0.576 6.93 1.38 12.48 0.014 −0.16 −2.37 2.04 0.885
Periurban vs. Rural** 2 0.22 −0.81 1.25 0.675 −0.09 −2.11 1.92 0.930 0.18 −0.52 0.88 0.620 0.80 −0.47 2.06 0.219 3.19 1.45 4.93 <0.001 −0.31 −1.02 0.41 0.400
3 0.24 −0.51 0.98 0.534 −0.44 −2.03 1.15 0.588 0.29 −0.23 0.81 0.275 0.60 −0.23 1.42 0.158 1.52 0.08 2.96 0.039 −0.05 −0.55 0.45 0.835
4 0.25 −0.52 1.02 0.522 −0.79 −2.32 0.74 0.311 0.40 −0.06 0.86 0.087 0.40 −0.37 1.16 0.309 −0.15 −1.57 1.26 0.830 0.20 −0.24 0.64 0.370
5 0.27 −0.82 1.36 0.630 −1.14 −3.00 0.72 0.230 0.52 −0.04 1.08 0.071 0.20 −0.94 1.34 0.735 −1.83 −3.51 −0.15 0.033 0.45 −0.12 1.02 0.121
6 0.28 −1.24 1.81 0.715 −1.49 −3.91 0.94 0.229 0.63 −0.13 1.39 0.104 0.00 −1.69 1.68 0.997 −3.50 −5.63 −1.37 0.001 0.70 −0.11 1.52 0.089
7 0.30 −1.70 2.30 0.769 −1.84 −4.94 1.26 0.246 0.74 −0.26 1.74 0.145 −0.20 −2.48 2.08 0.862 −5.17 −7.85 −2.50 <0.001 0.96 −0.13 2.05 0.086
8 0.32 −2.18 2.82 0.804 −2.19 −6.02 1.65 0.263 0.86 −0.40 2.11 0.182 −0.40 −3.29 2.49 0.785 −6.85 −10.11 −3.58 <0.001 1.21 −0.17 2.59 0.087
9 0.33 −2.67 3.34 0.828 −2.54 −7.12 2.05 0.279 0.97 −0.55 2.49 0.212 −0.60 −4.12 2.91 0.737 −8.52 −12.40 −4.64 <0.001 1.46 −0.22 3.14 0.089
Urban vs. Rural** 2 0.75 −0.23 1.73 0.135 −0.21 −2.09 1.67 0.824 0.47 −0.20 1.13 0.173 1.62 0.27 2.98 0.019 2.60 0.57 4.63 0.012 0.29 −0.47 1.05 0.459
3 0.65 −0.07 1.37 0.077 −0.15 −1.66 1.36 0.846 0.43 −0.08 0.93 0.098 1.07 0.19 1.96 0.017 1.91 0.23 3.58 0.026 0.18 −0.36 0.71 0.515
4 0.55 −0.22 1.32 0.161 −0.09 −1.58 1.41 0.910 0.39 −0.07 0.84 0.098 0.53 −0.30 1.35 0.213 1.21 −0.41 2.83 0.144 0.07 −0.40 0.53 0.780
5 0.45 −0.64 1.53 0.417 −0.02 −1.85 1.81 0.981 0.35 −0.21 0.91 0.225 −0.02 −1.26 1.22 0.971 0.51 −1.39 2.42 0.597 −0.04 −0.66 0.57 0.887
6 0.35 −1.16 1.86 0.651 0.04 −2.34 2.42 0.973 0.31 −0.45 1.06 0.423 −0.57 −2.40 1.26 0.540 −0.18 −2.59 2.22 0.882 −0.16 −1.03 0.72 0.728
7 0.25 −1.73 2.22 0.806 0.10 −2.93 3.14 0.946 0.27 −0.72 1.25 0.593 −1.12 −3.59 1.35 0.374 −0.88 −3.90 2.14 0.569 −0.27 −1.44 0.91 0.657
8 0.15 −2.31 2.60 0.906 0.17 −3.56 3.90 0.93 0.23 −1.00 1.46 0.716 −1.67 −4.81 1.47 0.297 −1.57 −5.26 2.11 0.403 −0.38 −1.87 1.11 0.621
9 0.05 −2.90 3.00 0.975 0.23 −4.22 4.68 0.919 0.19 −1.30 1.68 0.803 −2.22 −6.03 1.59 0.254 −2.27 −6.66 2.12 0.311 −0.49 −2.30 1.33 0.599
Open in a new tab

BMI, Body mass index; Est, Estimates of the difference in weight, height, or BMI between categorical groups; LL, Lower limit of 95% confidence interval; UL, Upper limit of 95% confidence interval; P, Probability estimate.

†

Results from two-level multilevel models adjusted for child age spline terms, physical activity (low tertile: referent), total energy intake, breastfeeding status (never breastfed: referent), maternal age group (24-39 years old: referent), maternal years of education, maternal height, maternal BMI, urbanicity (rural: referent), socioeconomic resources (low asset quartile: referent), and total annual income (< $10,000 tala: referent).

‡

Results from two-level multilevel models adjusted for child age spline terms, modern dietary pattern adherence (low: referent), total energy intake, breastfeeding status (never breastfed: referent), maternal age group (24-39 years old: referent), maternal years of education, maternal height, maternal BMI, urbanicity (rural: referent), socioeconomic resources (low asset quartile: referent), and total annual income (< $10,000 tala: referent).

§

Results from two-level multilevel models adjusted for child age spline terms, maternal age group, maternal education, urbanicity (rural: referent), and household total annual income (< $10,000 tala: referent).

**

Results from two-level multilevel models adjusted for child age spline terms, household total annual income (< $10,000 tala: referent).

Associations of household socioeconomic resources with body size trajectories

High or very high household assets were associated with higher BMI and velocity compared to low assets (Figures 1–2). Between ages 5-9 years, children from households with very high assets experienced an increasing BMI that crossed the 90th centile, suggesting many had overweight or obesity (BMI z-score > +1 SD). Whereas children with low or medium assets had a BMI that remained between the 50th and 75th centile for this same age period and were not classified as having overweight or obesity. Shapes of the body size trajectories illustrate greater gains in weight between 2-9 years for females with very high compared to low assets (p-value for interaction with age spline term 1 = 0.222 and age spline term 2 = <0.001) and this was similar in males (p-value for interaction with spline age term 1 = 0.312 and spline age term 2 = 0.001) (Supplementary Table 3). Although females with very high assets tended to have a higher height than those with low assets (Supplementary Figure 4), a lower height growth velocity between ages 2 and 9 years was associated with very high assets (p-value for interaction with spline age term 1 =0.014). There were no associations with height observed among males.

Associations of urbanicity with body size trajectories

Urbanicity was associated with higher BMI between ages 2 and 9 years (Figures 1–2). Between ages 2-9 years, females in the urban AUA and periurban NWU region had consistently higher BMI for age relative to the WHO standards and references than those in the rural ROU region. Between ages 5-9 years, males in the urban AUA region experienced a substantial increase in BMI from the 75th to 95th centile, and compared to the rural ROU region, the shapes of the trajectories appeared different with greater gains in BMI (p-value for interaction with age spline term 1=0.520 and age spline term 2= 0.014, Supplementary Table 4), and increasing weight that crossed from 50th to 90th centile (Supplementary Figure 5). Males living in the periurban NWU versus rural ROU region experienced a lower height growth velocity (p-values for interaction with age spline term 1 <0.001) and lower height by age 9 (95% CI: −12.40 - −4.64 cm, Supplementary Table 4), but also reached the 50th centile suggesting a relatively similar height to the WHO reference (Supplementary Figure 6).

Discussion

This study is among the first to characterize longitudinal changes in BMI traits related to childhood behavioral and environmental risk factors for obesity in Samoa. On average, consistently high BMI and weight relative to the WHO child standard and reference groups were observed across all ages, highlighting the need for intervention strategies in childhood before the age of 9. High adherence to a modern dietary pattern, lower physical activity (in females), higher household assets (indicative of greater socioeconomic resources) and living in the urban and peri-urban regions (in males) were associated with increasing childhood BMI that resulted in crossing to higher centile bands, indicating a higher risk of overweight or obesity. These findings may be used to focus public health efforts in the Samoan setting, informing who should be targeted with early intervention and which behavioral risk factors should be prioritized.

The positive associations of the modern dietary pattern score with childhood weight and BMI suggest that greater consumption of foods that make up the modern Samoan dietary pattern in early childhood could be a potentially important target for any obesity prevention intervention. The findings build upon our previously reported evidence that consistent consumption of a modern diet among Samoan children promotes gains in BMI z-scores between ages 2-7 years (3, 4). This is also consistent with findings from the Pacific Island Family study in New Zealand, where eating a diet high in protein and dairy (both present in our modern dietary pattern) was associated with greater BMI at age 4 years and weight gain (34). As Samoa continues to undergo nutritional and economic transitions, dietary patterns are likely to continue to modernize with increasing access to a range of imported and processed foods, including highly refined sources of carbohydrates such as white rice, bread, and noodles (8). Some families in Samoa are, however, becoming increasingly cognizant of the health risks associated with a more ‘Western’ dietary pattern and are attempting to revert to a traditional, more health-conscious diet. When additional data are available in the cohort, further longitudinal investigation of the changes in dietary patterns during childhood and beyond will be informative for intervention design.

Increased physical activity is known to support energy balance and bone development, while preventing the accumulation or storage of excess fat (35–37). Based on our prior findings, we hypothesize that the positive associations between physical activity and childhood weight and BMI z-score trajectories may reflect greater lean or muscle mass in those who are more active. In a random subsample of 3-7-year-olds from the cohort who received dual-energy x-ray absorptiometry (DXA) scans to assess body composition, we observed lower percent total body fat and higher percent lean mass with increasing accelerometer-measured daily physical activity (9).

The positive association of asset ownership with childhood BMI aligns with our prior cross-sectional analyses (3), but is not consistent with research among children living in other Pacific settings. In Hawai’i, greater socioeconomic status was associated with lower odds of overweight or obesity based on neighborhood-level education among Samoan children at ages 5 to 8 years (38). In the Growing Up in Aotearoa New Zealand cohort of multicultural Pacific children, living in areas of high deprivation with lower neighborhood socioeconomic resources was associated with a relatively higher BMI at age 8 years compared to the WHO child reference group (39). This is likely explained by Samoa being relatively earlier in the process of economic development and modernization than both of these settings, such that obesity continues to be more prevalent among those in high socioeconomic resource groups (6).

Among females in our Samoan sample, we observed lower height growth velocity in the very high compared to low asset groups. These socioeconomic differences in female height are difficult to explain and may reflect a multitude of factors experienced before birth, such as the intrauterine characteristics, height, or nutrition of mothers during pregnancy (40, 41), during infancy before age 2 years, child-rearing, or perhaps resource allocation based on sex (42). On the other hand, greater gains in weight relative to height contributed to the observed positive associations with BMI for males. Among all children, the differences in BMI trajectories across asset groups are worrisome for the development of obesity and additional efforts to monitor and prevent obesity are needed in this group, especially in the context of rising socioeconomic inequalities (43).

Rapid urbanization, which varies across the census regions in Samoa, has increased economic diversification (with many households shifting away from single income sources toward multiple sources from a growing range of sectors and markets) and infrastructure that may account for shifts in lifestyle and ownership of assets that drive increases in body size over time and subsequent obesity risk in children. Based on the latest 2016 census of the Samoan population, the highest labor force participation was observed in the urban AUA region, with lower rates of employment in agriculture, livestock, forestry, and fishery-related occupations compared to rural ROU (11.7% vs. 62%) (44, 45). As more family members in the urban AUA region engage in professional and clerical work than in the ROU region (20), we may anticipate greater household income and gains in consumer durables which could encourage poor eating behaviors and changes in daily routines of the children that influence growth and health. With a greater proportion of households owning a motor vehicle in urban and periurban regions (28-37%) than rural (18%)(46), the more urban settings may not likely be conducive for positive health behaviors if busier roads with more cars discourage physical activities and require further investigation.

The findings of this study should be interpreted in the context of its limitations. Convenience sampling to balance the regional distribution of the sample rather than to represent the national child population may limit generalizability. Exposure assessment occurred at similar age periods as the physical measurements of weight, height, and BMI (particularly 2-7 years), which precludes making causal inferences. Maternal-reported data are subject to recall bias, social desirability bias, and may depend on her involvement in the care of her child (47). We were limited to the available cohort data for this analysis and appreciate that additional information, including genetic data, documentation of participation in national health promotion programs, and/or family chronic disease history may be important to increase our understanding of childhood growth and development in Samoans. There are also multiple additional dimensions of socioeconomic position, wealth, and the environment that influence health and should be investigated in additional analyses (48). We also recognize that sex differences in height trajectories may be related to pubertal development, which was not measured in the cohort; however, the onset of menses and associated changes in body proportions is more likely to occur after the age of 9 years among Pacific children (49,50). Importantly, interpreting BMI in children has clinical limitations. Higher BMI may not equate to greater levels of body fat mass relative to lean mass and it is possible for risk factors to disproportionately influence BMI at the upper end of its distribution (51). In further studies, it will be important to explore the associations of these obesity risk factors with blood pressure, diabetes risk, and other cardiometabolic outcomes.

An important strength of our work, however, is the comprehensive data collected prospectively which allowed for the use of multilevel modeling. While the serial measurements of weight, height, and BMI vary among children in number and timing, the values were predicted by the study models using all available data. We were able to account for the non-linear shape of BMI trajectories in children (52, 53) using cubic spline terms for age. Considering the ongoing health transitions with changes in patterns of behaviors, morbidity, and mortality, Samoa is a unique population to continue studying the association between modifiable behaviors and environmental factors with growth, development, and health across the life course. With additional time points, data from the Ola Tuputupua’e cohort will enhance our understanding of etiologic pathways for obesity to concretely inform intervention strategies for Samoans and more broadly, Pacific people.

Conclusion

The observed inequalities in childhood body size trajectories – with high modern dietary pattern adherence, greater asset ownership, more urbanized regions, associated with high, centile-crossing BMI trajectories—suggest that health promotion programs and preventative interventions may be needed before the age of 9 years to address these risk factors. Interventions should focus on addressing multiple risk factors, reducing the consumption of the imported and micronutrient-poor foods that make up the modern diet in Samoa, and reinforcing environments to encourage positive health behaviors starting in early childhood.

Supplementary Material

Supplementary Materials

Acknowledgments

Thank you to the children and families who participate, as well as our partners in the Samoa Ministry of Health, Bureau of Statistics, Ministry of Women, Community, Social Development, and the OLaGA field team for making this work possible.

The Ola Tuputupua’e study received financial support from the following sources: Yale University (Faculty Funding, David Dull Internship Fund, Jan A.J. Stolwijk Fellowship Fund, Downs International Health Student Travel Fellowship, Thomas C. Barry Travel Fellowship), US National Institutes of Health (NIH) Minority and Health Disparities International Research Training Program (NIMHD T37MD008655), U.S. Fulbright Graduate Student Research Fellowship, Brown University (International Health Institute, Nora Kahn Piore Award, and Framework in Global Health Program), Brown University Population Studies and Training Center which receives funding from the NIH for training (T32 HD007338) and general support (P2C HD041020), and NIH National Lung, Health, Blood Institute for infrastructure support (R01 HL093093 and HL140570). WJ was supported by a UK Medical Research Council New Investigator Research Grant (MR/P023347/1) and acknowledges support from the National Institute for Health Research (NIHR) Leicester Biomedical Research Centre, which is a partnership between University Hospitals of Leicester NHS Trust, Loughborough University, and the University of Leicester. CCC was supported by Yale-Brown Ivy Plus Exchange Program, Ruth L. Kirschstein Predoctoral Individual National Research Service Award (NIH 1F31HL147414), Fogarty Global Health Equity Scholars Program (FIC D43TW010540), and now, NIH Pathway to Independence (K99HL166781).

Footnotes

Competing Interests

The authors declare no competing financial interest.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request and with the permission of the Health Research Committee of the Samoa Ministry of Health.

References

  • 1.United Nations Children’s Fund, World Health Organization, and The World Bank Group joint child malnutrition estimates: levels and trends in child malnutrition: key findings of the 2020 edition. 2020. [Google Scholar]
  • 2.NCD Risk Factor Collaboration. Heterogeneous contributions of change in population distribution of body mass index to change in obesity and underweight. Elife. 2021;10:e60060. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3.Choy CC, Desai MM, Park JJ, Frame EA, Thompson AA, Naseri T, et al. Child, maternal and household-level correlates of nutritional status: a cross-sectional study among young Samoan children. Public Health Nutr. 2017;20(7):1235–47. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Choy CC, Wang D, Naseri T, Soti-Ulberg C, Reupena MS, Duckham RL, et al. Longitudinal Assessment of Childhood Dietary Patterns: Associations with Body Mass Index z-Score among Children in the Samoan Ola Tuputupua’e (Growing Up) Cohort. Childhood Obesity. 2020. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.Choy CC, Lupematasila VF, Arorae MS, Tafunaina F, Unasa F, Soti-Ulberg C, et al. Prevalence of malnutrition among Samoan children aged 5 to 11 years in 2019–2020. Annals of Human Biology. 2023;50(1):200–5. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Choy CC, Hawley NL, Naseri T, Reupena MS, McGarvey ST. Associations between socioeconomic resources and adiposity traits in adults: Evidence from Samoa. SSM-Population Health. 2020:100556. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.Choy CC, Wang D, Baylin A, Soti-Ulberg C, Naseri T, Reupena MS, et al. Dietary patterns are associated with child, maternal and household-level characteristics and overweight/obesity among young Samoan children. Public Health Nutr. 2018;21(7):1243–54. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Seiden A, Hawley NL, Schulz D, Raifman S, McGarvey ST. Long-term trends in food availability, food prices, and obesity in Samoa. American journal of human biology : the official journal of the Human Biology Council. 2012;24(3):286–95. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Thompson AA, Duckham RL, Desai MM, Choy CC, Sherar LB, Naseri T, et al. Sex differences in the associations of physical activity and macronutrient intake with child body composition: A cross-sectional study of 3-to 7-year-olds in Samoa. Pediatric obesity. 2020;15(4):e12603. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Cole TJ, Faith MS, Pietrobelli A, Heo M. What is the best measure of adiposity change in growing children: BMI, BMI%, BMI z-score or BMI centile? European journal of clinical nutrition. 2005;59(3):419. [DOI] [PubMed] [Google Scholar]
  • 11.Wang D, Hawley NL, Thompson AA, Lameko V, Reupena MS, McGarvey ST, et al. Dietary Patterns Are Associated with Metabolic Outcomes among Adult Samoans in a Cross-Sectional Study. J Nutr. 2017;147(4):628–35. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.DiBello JR, McGarvey ST, Kraft P, Goldberg R, Campos H, Quested C, et al. Dietary patterns are associated with metabolic syndrome in adult Samoans. J Nutr. 2009;139(10):1933–43. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Baylin A, Deka R, Tuitele J, Viali S, Weeks DE, McGarvey ST. INSIG2 variants, dietary patterns and metabolic risk in Samoa. Eur J Clin Nutr. 2013;67(1):101–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Miki T, Eguchi M, Akter S, Kochi T, Kuwahara K, Kashino I, et al. Longitudinal adherence to a dietary pattern and risk of depressive symptoms: The Furukawa Nutrition and Health Study. Nutrition. 2018;48:48–54. [DOI] [PubMed] [Google Scholar]
  • 15.Bielemann RM, Reichert FF, Paniz VM, Gigante DP. Validation of the Netherlands physical activity questionnaire in Brazilian children. International Journal of Behavioral Nutrition and Physical Activity. 2011;8(1):45. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.Damasceno VdO, Rabelo AS, Lamounier JA, Szmuchrowski LA, Couto BP, Souza DEd, et al. Criteria validity of the Netherlands physical activity questionnaire for children. Revista Brasileira de Cineantropometria & Desempenho Humano. 2018;20(6):504–14. [Google Scholar]
  • 17.Janz KF, Broffitt B, Levy SM. Validation evidence for the Netherlands physical activity questionnaire for young children: the Iowa bone development study. Research quarterly for exercise and sport. 2005;76(3):363–9. [DOI] [PubMed] [Google Scholar]
  • 18.Warmath C, Choy CC, Frame E, Sherar LB, Duckham RL, Soti-Ulberg C, Naseri T, Reupena MS, Hawley NL. Concordance of the Netherlands Physical Activity Questionnaire with accelerometer-measured physical activity in Samoan children. 2021. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Samoa Bureau of Statistics. Population and Housing Census 2011 Analytical Report. Apia, Samoa; 2012. Available from: http://sdd.spc.int/en/resources/document-library?view=preview&format=raw&fileId=223. [Google Scholar]
  • 20.Samoa Bureau of Statistics. 2016 Census Brief No. 1: Revised version population snapshot and household highlights. Apia, Samoa; 2017. [Google Scholar]
  • 21.Greenland S, Pearl J, Robins JM. Causal diagrams for epidemiologic research. Epidemiology. 1999:37–48. [PubMed] [Google Scholar]
  • 22.Haddad J, Ullah S, Bell L, Leslie E, Magarey A. The Influence of Home and School Environments on Children’s Diet and Physical Activity, and Body Mass Index: A Structural Equation Modelling Approach. Matern Child Health J. 2018;22(3):364–75. [DOI] [PubMed] [Google Scholar]
  • 23.Hendrie GA, Coveney J, Cox DN. Defining the complexity of childhood obesity and related behaviours within the family environment using structural equation modelling. Public Health Nutr. 2012;15(1):48–57. [DOI] [PubMed] [Google Scholar]
  • 24.Gilliland JA, Rangel CY, Healy MA, Tucker P, Loebach JE, Hess PM, et al. Linking childhood obesity to the built environment: a multi-level analysis of home and school neighbourhood factors associated with body mass index. Can J Public Health. 2012;103(9 Suppl 3):eS15–21. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Ong JX, Ullah S, Magarey A, Leslie E. Positive influences of home food environment on primary-school children’s diet and weight status: a structural equation model approach. Public Health Nutr. 2016;19(14):2525–34. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26.Santiago-Torres M, Cui Y, Adams AK, Allen DB, Carrel AL, Guo JY, et al. Structural equation modeling of the associations between the home environment and obesity-related cardiovascular fitness and insulin resistance among Hispanic children. Appetite. 2016;101:23–30. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.Popkin BM. Urbanization, lifestyle changes and the nutrition transition. World development. 1999;27(11):1905–16. [Google Scholar]
  • 28.Neuman M, Kawachi I, Gortmaker S, Subramanian SV. Urban-rural differences in BMI in low- and middle-income countries: the role of socioeconomic status. Am J Clin Nutr. 2013;97(2):428–36. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.Pirgon Ö, Aslan N. The role of urbanization in childhood obesity. Journal of clinical research in pediatric endocrinology. 2015;7(3):163. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.SAS Institute. SAS/STAT 14.2 User’s Guide. 2016. [Google Scholar]
  • 31.Choy CC, Johnson W, Braun JM, Soti-Ulberg C, Reupena MS, Naseri T, et al. Associations of childhood BMI traits with blood pressure and glycated haemoglobin in 6–9-year-old Samoan children. Pediatric Obesity. 2024:e13112. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Amrhein V, Greenland S, McShane B. Scientists rise up against statistical significance. Nature Publishing Group; 2019. [DOI] [PubMed] [Google Scholar]
  • 33.Md Onis, Onyango AW, Borghi E, Siyam A, Nishida C, Siekmann J. Development of a WHO growth reference for school-aged children and adolescents. Bulletin of the World health Organization. 2007;85:660–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34.Rush E, Paterson J, Obolonkin V. Food frequency information—relationships to body composition and apparent growth in 4-year-old children in the Pacific Island Family Study. Clinical Correspondence. 2008. [PubMed] [Google Scholar]
  • 35.García-Hermoso A, Saavedra J, Ramírez-Vélez R, Ekelund U, del Pozo-Cruz B. Reallocating sedentary time to moderate-to-vigorous physical activity but not to light-intensity physical activity is effective to reduce adiposity among youths: a systematic review and meta-analysis. Obesity Reviews. 2017;18(9):1088–95. [DOI] [PubMed] [Google Scholar]
  • 36.Janssen I, LeBlanc AG. Systematic review of the health benefits of physical activity and fitness in school-aged children and youth. International journal of behavioral nutrition and physical activity. 2010;7:1–16. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37.Kohrt WM, Bloomfield SA, Little KD, Nelson ME, Yingling VR. Physical activity and bone health. Medicine & Science in Sports & Exercise. 2004;36(11):1985–96. [DOI] [PubMed] [Google Scholar]
  • 38.Novotny R, Oshiro CES, Wilkens LR. Prevalence of childhood obesity among young multiethnic children from a health maintenance organization in Hawaii. Childhood Obesity. 2013;9(1):35–42. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39.Morton SMB, Walker CG, Gerritsen S, Smith A, Cha J, Atatoa Carr P, et al. Growing Up in New Zealand: A longitudinal study of New Zealand children and their families. Now We Are Eight. Auckland: Growing Up in New Zealand; 2020. [Google Scholar]
  • 40.Addo OY, Stein AD, Fall CH, Gigante DP, Guntupalli AM, Horta BL, et al. Maternal height and child growth patterns. The Journal of pediatrics. 2013;163(2):549–54.e1. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 41.Howe LD, Tilling K, Galobardes B, Smith GD, Gunnell D, Lawlor DA. Socioeconomic differences in childhood growth trajectories: at what age do height inequalities emerge? J Epidemiol Community Health. 2012;66(2):143–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 42.Thomas D. Like father, like son; like mother, like daughter: Parental resources and child height. Journal of human resources. 1994:950–88. [Google Scholar]
  • 43.Australian Government AusAID. Samoa Country Case Study. Canberra, Australia; 2012. [Google Scholar]
  • 44.Samoa Bureau of Statistics. Samoa socio-economic atlas 2016. Apia, Samoa; 2016. [Google Scholar]
  • 45.Samoa Bureau of Statistics. 2016 Census Brief No. 3 Education and Economic Characteristics. Apia, Samoa; 2018. [Google Scholar]
  • 46.Samoa Bureau of Statistics. 2016 Census Brief No. 4 Housing/Dwelling Details & Household Amenities and Capital Goods. Apia, Samoa; 2018. [Google Scholar]
  • 47.Sallis JF, Saelens BE. Assessment of physical activity by self-report: status, limitations, and future directions. Research quarterly for exercise and sport. 2000;71(sup2):1–14. [DOI] [PubMed] [Google Scholar]
  • 48.Hruschka DJ, Hadley C, Hackman J. Material wealth in 3D: Mapping multiple paths to prosperity in low-and middle-income countries. PloS one. 2017;12(9):e0184616. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 49.Rush E, Tautolo el- S, Paterson J, Obolonkin V. Pacific Islands Families Study: Signs of puberty are associated with physical growth at ages 9 and 11 years. N Z Med J. 2015. Nov 20;128(1425):24–33. [PubMed] [Google Scholar]
  • 50.Kubo A, Acker J, Aghaee S, Kushi LH, Quensenberry CP, Greenspan LC et al. Pubertal Timing Across Asian American, Native Hawaiian, and Pacific Islander Subgroups. JAMA Netw Open. 2024;7(5):e2410253. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 51.Johnson W, Li L, Kuh D, Hardy R. How has the age-related process of overweight or obesity development changed over time? Co-ordinated analyses of individual participant data from five United Kingdom birth cohorts. PLoS medicine. 2015;12(5):e1001828. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 52.Cameron N, Bogin B. Human growth and development: Academic Press; 2012. [Google Scholar]
  • 53.Cole T. Children grow and horses race: is the adiposity rebound a critical period for later obesity? BMC pediatrics. 2004;4(1):6. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Materials
NIHMS2035568-supplement-Supplementary_Materials.pdf (1,004KB, pdf)

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request and with the permission of the Health Research Committee of the Samoa Ministry of Health.

RESOURCES